Datasets

xSDL dataset

This dataset contains data from 10 participants (4 female) from 30 to 59 years old (mean 36.2). All participants had normal vision without the need of any correction lenses.

Each participant looked at 35 circular targets arranged in a 5x7 grid. Data was collected simultaneously using a split mirror and two Sony PS3 cameras:

  1. Camara 1 used active near IR illumination (Stroboscopic Differential Lighting)
  2. Camera 2 used an off-axis near IR illumination

More details in the paper: "xSDL - Stroboscopic Differential Lighting Eye Tracker with Extended Temporal Support".


Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Download: https://latin.ime.usp.br/static/data/xSDL_data.tar.gz

SLD Framework dataset

This dataset contains eye images from 7 subjects, collected for both left and right eyes at 9 distinct head positions. In each position, each subject gazed at 49 screen targets arranged in a 7x7 grid. For each screen target being gazed, 20 samples were acquired. Note that capture for the left and right eyes were not carried out simultaneously.

This dataset was used to perform the evaluation of gaze estimation methods presented in the paper "Screen-Light Decomposition Framework for Point-of-Gaze Estimation Using a Single Uncalibrated Camera and Multiple Light Sources". It is the same dataset used in the previous paper "Improving Head Movement Tolerance of Cross-Ratio Based Eye Trackers", in which detailed information about how the dataset was generated can be found.


Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.

Download: https://latin.ime.usp.br/static/data/SLD_Framework_dataset.tar

Pursuit Dataset 2020

This dataset has been collected for evaluation of pursuit detection algorithms into four distinct combinations of shape (circle and square) and velocity (120°/s and 240°/s) of the pursued object trajectory. 16 participants performed two sessions composed by four pursuit tasks and two non-pursuit tasks. The data were collected through a 200Hz binocular eye tracker as sequences of points that correspond to the pupil center of each eye.

Download: https://latin.ime.usp.br/static/data/pursuit_dataset.zip